INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

boards

-0.75

 Birch

-0.74

Story

-0.73

fx

-0.72

umbo

-0.68

achy

-0.67

zin

-0.66

hander

-0.66

 Score

-0.66

ylan

-0.64

POSITIVE LOGITS

BIP

0.76

 nomine

0.75

æ©

0.75

ividual

0.72

NetMessage

0.72

ioch

0.71

soever

0.70

iferation

0.70

absolute

0.69

 Plasma

0.68

Activations Density 0.000%

No Known Activations

This feature has no known activations.